phase11: merge 45 + cookbook 154 — 581 bodies / 590 regions

Worker A's 0x800F452C (112 B, maspsx=epilogue, second attempt) -- a sibling of its own
0x800F4098, matched with the mid-function-branch-target lever applied first try.

Worker A asked FIVE TIMES for the lui-page rule to be prominent, having made the mistake five
times. It is now cookbook 154 and the FIRST of the five things the workflow doc tells a new
worker, with all five instances tabulated. The fifth instance added the missing half: the
immediate is a SIGNED DISPLACEMENT, and the signature of the error is 'a residual of ONE
differing byte whose only difference is a lui or addiu immediate'.

Also confirmed from worker A's side: the shape A/B correction works -- 0x800F452C is a shape-B
row and closed with the single token, so the tail filter is obsolete and A has dropped it.

config/regions.tsv

@@ -484,6 +484,7 @@
 0x800F4098	0x800F4100	src/func_800F4098.c	maspsx=epilogue
 0x800F42AC	0x800F430C	src/func_800F42AC.c	maspsx=epilogue
 0x800F44D0	0x800F452C	src/func_800F44D0.c	maspsx=epilogue
+0x800F452C	0x800F459C	src/func_800F452C.c	maspsx=epilogue
 0x800F5200	0x800F521C	src/func_800F5200.c
 0x800F5AF8	0x800F5B40	src/func_800F5AF8.c
 0x800F5B40	0x800F5B70	src/func_800F5B40.c

docs/MATCHING_COOKBOOK.md

@@ -2525,3 +2525,30 @@ by a worker who has not seen it:
 
 Worker D's `0x800FAF84` writes through `sp+16` with the saved `s0` at `sp+24`, so the local is
 **8 bytes** while the callee writes past it. **That is a real property of the original.**
+
+### 154. *** READ THIS FIRST: derive every absolute address from the (page, signed immediate) PAIR ***
+
+**Worker A asked for this line five times, having made the mistake five times. It is the single most
+repeated error in this project, so it goes first.**
+
+    lui  a0,0x8011
+    addiu a0,a0,-6504      <- the address is 0x80110000 - 6504 = 0x8010E698
+                           <- NOT 0x8011E698, and NOT 0x8010E6A8
+
+**The `lui` immediate is the PAGE. The second instruction's immediate is a SIGNED DISPLACEMENT from
+that page.** Reading the printed value as "the low half of the address" is wrong, and reading the
+magnitude without its sign is wrong in the same way.
+
+**THE SIGNATURE IS DIAGNOSTIC:** *a residual of ONE differing byte whose only difference is a `lui`
+or `addiu` immediate.* If you see that, you have made this error — do not go looking for a source
+shape. Worker A's five instances:
+
+| row | error | cause |
+|---|---|---|
+| gp-relative | `sw zero,2132(gp)` read as the wrong global | offset guessed instead of derived from `_gp = 0x80121938` |
+| `0x800F44D0` | `lui v0,0x8012; lw v0,-22316(v0)` read as 0x8012A8D4 | the page is not part of the address; correct is **0x8011A8D4** |
+| `0x800F42AC` | read as 0x8011E644 | correct is **0x8010E644** |
+| `0x800F4098` | read as 0x8011E5F4 | correct is **0x8010E5F4** |
+| `0x800F452C` | read as 0x8010E6A8 | **the sign**: correct is **0x8010E698** |
+
+**Derive it arithmetically, every time: `page * 0x10000 + sign_extend(immediate)`.**

docs/ORCHESTRATOR_WORKFLOW.md

@@ -436,3 +436,23 @@ so the ledger is session-scoped by design — it coordinates live workers, not h
 assumption the work assignment rests on.** Adjacency optimised *which* row to take without anyone
 revisiting *who owns it*. When you adopt a worker-discovered heuristic, re-check what it assumes
 about your own process.
+
+
+## The five things to tell a new worker FIRST (Phase 11)
+
+Ordered by how much time each has cost workers in practice:
+
+1. **Derive every absolute address from the `(page, signed immediate)` PAIR** — `lui a0,0x8011` /
+   `addiu a0,a0,-6504` is `0x80110000 - 6504 = 0x8010E698`, not `0x8011E698` and not `0x8010E6A8`.
+   **The signature of getting this wrong is a residual of ONE differing byte whose only difference is
+   a `lui` or `addiu` immediate.** This was the most repeated error in the phase — five instances,
+   all one worker.
+2. **A match is `differing_bytes=0` AND `result=MATCH`,** and the whole-binary gate must stay green.
+   Functional equivalence is not a match.
+3. **Cost is tie-break density, not size.** 548 B matched first attempt; 176 B took nine and never
+   closed. **Prefer a repetitive body over a small one**, and prefer small among equal redundancy
+   because the milestone counts BODIES.
+4. **When the structure is confirmed and only register numbers differ, CLASSIFY — do not grind.**
+   That is the allocator class, and it is where most failures live.
+5. **Check the row is free with a CONTAINMENT test** (a region's end address is exclusive), **append
+   `wip` before starting, and append `released` only for a row you yourself took.**